CA Oil Spills 2008

Overview:

This code creates maps depicting location and number of oil spills in each county in California in 2008. A G-function analysis was carried out to asses whether clustering is present in the data. See code for annotations.

Exploratory Interactive Map

First, lets see what our data looks like using a basic interactive tmap:

Show code

# Create basic interactive exploratory plot using tmap
tmap_mode(mode = 'view')
tm_shape(ca_sf) +
  tm_borders(col = 'black') +
  tm_shape(oil_spill_sf) +
  tm_dots()

Figure 1. Black points represent oil spills in California in 2008.

Choropleth Map: Oil Spill Incidents per County

Make a map that colors counties based on the relative number of oil spills that occurred in each county in 2008.

Show code

# Join oil spill and California counties data sets/geometry so we have themn all in one place
ca_oil_sf <- ca_sf %>%
  st_join(oil_spill_sf)

# Count number of oil spills per county (exclude counties that have NA values)
oil_counts_sf <- ca_oil_sf %>%
  group_by(county_name) %>%
  summarize(n_records = sum(!is.na(county_name))) 

# Create a choropleth map using geom_sf()
ggplot(data = oil_counts_sf) +
  geom_sf(aes(fill = n_records), color = 'white', size = 0.1) +
  scale_fill_gradientn(colors = c('lightgrey', 'darkorchid', 'navyblue')) +
  theme_void() +
  labs(fill = "Number of Oil Spills",
       title = "CA Oil Spills Per County 2008")

Figure 2. Oil spills per county in California in 2008. Darker shades of purple represent larger numbers of oil spills.

Show code

# Just for fun, make a cool interactive choropleth map using tmap
tmap_mode(mode = 'view')
tm_shape(oil_counts_sf) +
  tm_borders(col = 'black') +
  tm_fill('n_records', palette = 'BuPu', title = "Number of Oil Spills")

Figure 3. Same map as above, but interactive: Oil spills per county in California in 2008. Darker shades of purple represent larger numbers of oil spills.

Nearest Neighbor G-Function Analysis

Lets say we want to find out whether or not these oil spill incidents are clustered or random. We can find out using a G-function analysis.

Show code

oil_sp <- as(oil_spill_sf, 'Spatial') # Convert to object 'Spatial'
oil_ppp <- as(oil_sp, 'ppp') # Convert to spatial point pattern

ca_sp <- as(ca_sf, 'Spatial') # Convert to object 'Spatial'
ca_win <- as(ca_sp, 'owin') # this window will exclude marine oil spills, but there is still enough data to perform G Function analysis without these observations

# Combine as a point pattern object (points + window):
oil_full <- ppp(oil_ppp$x, oil_ppp$y, window = ca_win) 

# make vector containing values from 0 - 10,000 that will be used to calculate G(r)
r_vec <- seq(0, 10000, by = 100) 

gfunction <- envelope(oil_full, fun = Gest, r = r_vec, nsim = 20, nrank = 2) # Calculate the actual and theoretical G(r) values, using 20 simulations of CRS for the "theoretical" outcome (the processing time is very low so fewer simulations will run faster)

Generating 20 simulations of CSR  ...
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,  20.

Done.

Show code

# Create dataframe to be used for G-function plot
gfunction_long <- gfunction %>%
  as.data.frame() %>%
  pivot_longer(cols = obs:hi, names_to = 'model', values_to = 'g_val')

Show code

# Create basic graph of G-function analysis
ggplot(data = gfunction_long, aes(x = r, y = g_val, group = model)) +
  geom_line(aes(color = model)) +
  labs(title = "G-function Plot", group = "Model")

Figure 3. G Function Analysis.

This confirms clustering - our data has a greater proportion of events with nearest neighbor at smaller distances compared to a theoretical CSR scenario.

Citation:

CA DFW Oil Spill Incident Tracking https://gis.data.ca.gov/datasets/7464e3d6f4924b50ad06e5a553d71086_0/data